The feature only works for English and it's been enabled for a small number of channels that usually feature talks and interviews: UC Berkeley, Stanford, MIT, Yale, UCLA, Duke,UCTV, Columbia, PBS, National Geographic.
Another new feature is auto-timing, which lets you upload the transcription of a video and it automatically generates the time codes. "All you need to do is create a simple text file with all the words in the video and we'll use Google's ASR technology to figure out when the words are spoken and create captions for your video."
Since Google's speech recognition technology is not perfect, it would be useful to generate the captions and then to manually edit them to correct the mistakes.
Automatic captions make YouTube videos more accessible: you can watch videos with the sound off and you can translate the captions into another language using Google Translate.